It is similar to a spell checker, while spell checkers improve spelling in general, they do not improve a persons spelling ability. Instead acting as a crutch. No need to spell well when the machine will do it for you.
Some people really like expanding their thoughts via LLM prompt. Some so much it acts like a big crutch, no need to think coherently, the machine will do it for you. So they use the LLM for everything.
As a related tangent something is messed up in my web browser spell checker, it gives the red squiggles indicating a misspelling, but refuses to give suggested corrections. I would fix it but... My spelling ability has never been better than it is right now.
Also, some people are weirdos.
I've lost count of how many times I've clicked into a bio from a flagged, obviously LLM written comment to find some variation of "Building new AI tools for agentic devops"...
So why use an LLM for commenting? See above, people are weird.
Time / quantity.
Pre-LLMs you needed a whole "marketing" agency to astroturf on a meaningful scale. Post-LLMs a single person can run multiple astroturfing campaigns in parallel.
For example, consider this prompt -- "Find topics that would be relevant to people interested in <company> and post a topical post that mentions <product>".
Also, influencing public opinion on certain topics, such as Israel or Palestine, the current administration, the democrats, the various wars that are ongoing, or AI itself.